Learn how to train a Generative Adversarial Network (GAN) using StyleGAN2 ADA with your own images in this comprehensive video tutorial. Discover the process of creating GANs from various image sources, including fish, holiday photos, and Minecraft. Follow step-by-step instructions on setting up a Docker image, converting images to TFRecords, and training the GAN. Gain insights into potential issues during image conversion and understand GAN training output. Explore examples of GANs generated from 1970s sci-fi magazine covers, NVIDIA FFHQ faces, and more. Acquire the knowledge to create unique GANs tailored to your specific image datasets.
Overview
Syllabus
Using a Lenovo ThinkStation P920 with Dual NVIDIA RTX 8000 GPUs
Planning your GAN Training Project
A GAN to Generate 1970s SciFi Magazine Covers
NVIDIA FFHQ Faces GAN
How I collected images for FishGAN
How I created Minecraft GAN
How I created Christmas GAN
Setting up the Docker Image
Starting Docker
Convert to TFRecords
What can go wrong converting your images
Training the GAN
Understanding GAN Training Output
Conclusions
Taught by
Jeff Heaton
Related Courses
-
Building Generative Adversarial Networks
1.0 -
Build Basic Generative Adversarial Networks (GANs)
-
Advanced Generative Adversarial Networks (GANs)
-
Apply Generative Adversarial Networks (GANs)
-
Training a GAN on Your Own Images - StyleGAN2 ADA PyTorch
-
Training StyleGAN2 ADA PyTorch Images with Low GPU Memory - NVIDIA GEFORCE RTX 3060 Ti
